Modelpublic

jindaznb/torgo_tiny_finetune_F03_frozen_encoder

sourceHugging Faceapache-2.0updated 2y agoView on Hugging Face

0likes1downloads

Model Card

torgotinyfinetuneF03frozen_encoder

This model is a fine-tuned version of openai/whisper-tiny on the None dataset. It achieves the following results on the evaluation set:

—Loss: 0.0487
—Wer: 34.9794

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

—learning_rate: 0.0001
—trainbatchsize: 16
—evalbatchsize: 1
—seed: 42
—optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
—lrschedulertype: linear
—lrschedulerwarmup_steps: 1000
—num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.7886	0.85	500	0.0571	14.9520
0.0987	1.69	1000	0.0536	50.3429
0.0695	2.54	1500	0.0480	4.3896
0.0479	3.39	2000	0.0534	7.9561
0.0314	4.24	2500	0.0542	5.0754
0.0239	5.08	3000	0.0438	5.0754
0.0173	5.93	3500	0.0399	7.8189
0.0122	6.78	4000	0.0402	7.4074
0.0099	7.63	4500	0.0384	5.0754
0.0091	8.47	5000	0.0380	4.6639
0.0077	9.32	5500	0.0400	9.6022
0.0057	10.17	6000	0.0361	8.0933
0.0043	11.02	6500	0.0377	15.9122
0.0028	11.86	7000	0.0338	15.6379
0.0026	12.71	7500	0.0407	16.7353
0.0025	13.56	8000	0.0404	16.3237
0.0022	14.41	8500	0.0387	13.3059
0.0014	15.25	9000	0.0373	19.4787
0.0012	16.1	9500	0.0414	25.2401
0.0006	16.95	10000	0.0475	28.3951
0.0004	17.8	10500	0.0435	30.3155
0.0004	18.64	11000	0.0480	32.0988
0.0002	19.49	11500	0.0487	34.9794

Framework versions

—Transformers 4.32.0
—Pytorch 2.1.0+cu121
—Datasets 2.14.7
—Tokenizers 0.13.3